In October 2023, Time magazine announced the best inventions of the year. Among the 14 selected applications in the field of AI, the veteran software company Adobe's Generative Fill surpassed OpenAI's GPT-4 and became the leading position in the segment. Time Magazine: Best Inventions of 2023 (AI Section).
Once Midjourney's AI-generated ** came out, it was once considered a "lifeline" for Photoshop. However, after a year of development, from the perspective of the market,The success of new products such as Midjourney and Stable Diffusion in the AIGC space has helped Adobe to quickly develop and implement the development of large AI models in applications.
Year-to-date, Adobe has shown a 71% increase in the U.S. marketIts market capitalization increased by a full $100 billionEquivalent to the market capitalization of a public company. Adobe's performance may be driven by the expectation of the growth of the business value of generative AI.
At present, the most concerned and earliest commercialized generative AI application field is AIGC, which can generate a variety of content such as text, images, audio**, and decision-making reference suggestions. Among them,AI painting tools and potential personal intelligent assistant AI robots are the two applications that have attracted the most attention and landed first. "AI painting is hot out of the circle, can big factories enter the game and bloom everywhere to roll out the new track of AIGC?".》。
In October of this year, at the annual creative conference, Adobe unveiled three new Firefly series AI models. According to the company's official statement, these models will be closely integrated with Adobe's products such as Photoshop, Premiere, After Effects, etc., greatly improving the production efficiency of users. The capital market's confidence in the commercialization potential of generative AI is also strong.
Shantanu Narayen, CEO of Adobe, said in an interview that while there are many venture-backed AI companies emerging, they will eventually face restructuring due to the lack of a clear business model. Dr. Luo Rentong, an expert in the application and practice of digital economy, said that the current global development of artificial intelligence is entering the early stage of the era of general artificial intelligence represented by large language models, and Adobe's actions and initiatives on the path of commercialization of generative AI provide a key demonstration for other companies.
In March 2023, Adobe released Firefly, which already had features such as text-generated images, AI-generated text effects, and recoloring, and later added features such as generative filling, text generation**, and posters.
One of Firefly's key innovations is that it solves copyright issues in the commercialization of generative AI。The training data for Firefly's large model comes from Adobe Stock, an image library that contains publicly licensed or expired copyrights. When a creator uploads a work to the gallery, a transaction is formed when other users** work it, and the creator also receives the corresponding royalty income.
In October, Adobe released an iteration of the Firefly model and added more than 100 AI feature updates to Creative Cloud, such as smart enrichment** in Photoshop. Make AI-generated ** better integrated into the workflow of Adobe Family Bucket.
Previously, AI-generated ** different prompt words will produce huge differences, resulting in low cost performance to use. Because for users, the core of AI-generated ** lies in the luck of using prompt words, even if the same prompt word is entered every time, the generated ** may be completely different. Applications such as the Wonderful Duck Camera functionalize prompt words, which lowers the threshold for AIGC to use.
Another issue is that the generated ** must be edited when it is commercialized. Whether it's Midjourney or Stable Diffusion, after generating a **, the designer also needs to edit it in Photoshop, and even need to convert the ** into a vector image for easy editing.
That's why Adobe has created Firefly vector models, which allow AI to directly generate editable vector graphics. Adobe's Illustrator has also launched a functional test of text-generated vector graphics. This means that users can complete the entire process from image generation to editing in the Adobe family of products, and the migration cost is very low.
So, while apps like Midjourney, Dall-E, and others can be generated**, they are more focused on the build itself, with very limited editing capabilities and can't integrate with professional tools like Adobe Family Bucket. It's fair to say that Midjourney is more of an alternative to photo galleries like Flickr and Shutterstock.
In the professional commercialization scenario, Adobe is still an irreplaceable leader. Dr. Luo Rentong, an expert in the application and practice of the digital economy, believes that the importance of the AI model lies in the application practice, and only in this way can its value be reflected. In the same way, a model no matter how big it is, doesn't make sense. Although there is still a gap between Adobe's AI technology capabilities and similar competitors such as Midjourney in some details, it has clearly found the pain points and needs of the market, and used its own accumulation to give full play to the advantages of its own products.
In these business scenarios, Adobe's products cover a wide range of products. In addition to the familiar Photoshop, there is also InDesign for UI design, LightRoom for editing, and Illustrator for vector manipulation. Adobe has an absolute dominance in the graphic design segment, and its market competitiveness has been built up.
It can be said that Adobe has seized the market position in the market segment of graphic design, and the value of AI is the key to impressing the capital market, not just by virtue of its AI technology capabilities.
As the world's leading office software, Microsoft has also attracted the attention of the capital market. However, investors' eyes are not just on Microsoft's technical prowess, but also on the working people who pay for Microsoft 365 on time every year.
What Microsoft and Adobe have in common is the integration of large language models with existing software products, with the goal of simplifying the use of the software. Increase user productivity and productivity with more natural interactions, such as direct text prompts.
YetEven industry leaders like Adobe and Microsoft face a serious problem: the high cost of computing power. According to a report by The Information, Midjourney, a high-profile influencer company, has generated just $200 million in revenue this year. While this number is not small, it still falls far short of achieving the goal of commercializing AIGC.
For capital markets, Adobe's business model is almost perfect:Occupy a dominant market position, have a large group of paying users, high profit margins, and are highly compatible with AI mapping and computing. If such a company is unable to make significant gains in the field of artificial intelligence, it will undoubtedly have an impact on the prospects of AI graphics computing for the entire industry and reduce people's confidence in its expectations.
In simple terms,Adobe has devised a sophisticated approach to pricing for Firefly: Charges are charged on a point-based basis. To put it simply, one credit can be used to generate a **, and users can get 25 credits per month for free, with additional purchases if they need more. Users can purchase the Firefly service or CC Family Bucket separately, with the option to pay monthly or annually, and the benefits enjoyed by individuals and businesses are also different. In order to prevent users from losing money due to overuse, Adobe will slow down the service if a user uses more than the credits allocated each month.
A large part of the cost structure of most Internet products is relatively fixed operating costs, including cloud services. This part of the cost will gradually decrease as the scale of users increases. However, the situation of AIGC products is different, and every user interaction is computed once in the cloud, resulting in corresponding costs. The more users use it, the higher the cost. Developers can only reduce the computing power required for a single interaction through software optimization, but the cost of "use once" cannot be changed.
Although large language models are still in their infancy in terms of industry and commercialization, the market is still hot, and various large models are emerging one after another. According to public data, more than 130 large models of various types have been launched in China. "In-depth Inventory|Mobile phone manufacturers are all in the game!Huawei, Xiaomi, vivo large model mobile phones are unveiled, will it be the next explosion point of smartphones?》。
According to Dr. Luo Rentong, an expert in the application and practice of digital economy, the development of global artificial intelligence has entered the early stage of the era of general artificial intelligence, represented by large language models. It will take another 2 to 5 years for the industry to usher in a breakthrough period. On the one hand, models and algorithms require greater innovation and development;On the other hand, in the face of the development of multimodal large language models, more high-quality data resources and larger-scale AI computing resources are also needed.
According to a Bloomberg industry research report, the market size of generative AI will grow from $40 billion in 2022 to 1.1 billion in 2032 in the next 10 years$3 trillion. Generative AI is expected to account for 10% to 12% of IT hardware, software services, advertising spending, and gaming market spending by 2032, up from less than 1% today. As the U.S. tightens its controls on semiconductor exports to China, the pace of domestic AIGC development may be affected.
However, China has a huge number of users and a large consumer market, which provides a huge business prospect for the development of the domestic hardware side, which may give rise to its accelerated development. Therefore, for companies participating in the domestic AIGC industry, waiting for opportunities is the basic tone at present. The history of the Internet industry shows that it is often not the first companies to enter the market that end up being large companies. At present, the AIGC industry is still foggy and has not yet entered the formal rise stage, so it is important to "survive".
InnovationCases are welcomeLike and share!~~
Free eBook:Structured Thinking |Thorny business paths|Ten Words of the YearInnovative Enterprise Cases |Blockchain Finance
Luo Rentong, an expert in the application and practice of digital economy, gave a lecture
Smart Applications |"Internet of Energy"."New Power System"."Smart Building"."Smart Airport"."Smart Transportation".Tobacco Digitalization"Smart City"."Numbers**".
Digital Applications |"Big Data and Middle Platform".Blockchain"Metaverse".
Artificial Intelligence |Trends in Artificial IntelligenceChatGPT New Era
Digital Transformation |"Digitalization of Manufacturing".Digitalization in Retail
Finance Courses |Fintech InnovationReform of Commercial BanksDigital Yuan".
Thinking & Skills |Digital Management ThinkingDigital Organizational Transformation"User Thinking".
Macro & Trends |"2023 Macro Trends".Digital Economy"Dual Carbon Strategy"."New Energy Vehicles".
Innovation & Entrepreneurship |"Entrepreneurship Mentor Ability Improvement Training"."Empowerment of Small and Medium-sized Enterprises".